Just know what my tests revealed. Try it yourself and see.Mike, which part of the original post doesn't work for color sensors like Bayer?
Some time ago I tested the MTF of the SD9 and Bayer G9 where the color MTFs were shown in addition to the grayscale Y' response:
Best viewed original size.
Does anything there relate to MTF somehow "failing to work" for the Bayer?
Unfortunately, they're on a HD that died some years ago (along with all of my original SD10 and 14 raws, before I understood the importance of having a backup drive). But they were all 16 bit tiffs.
Gee thanks. I'll have to go and find "a flower" and see what happens.Just know what my tests revealed. Try it yourself and see.Mike, which part of the original post doesn't work for color sensors like Bayer?
Some time ago I tested the MTF of the SD9 and Bayer G9 where the color MTFs were shown in addition to the grayscale Y' response:
Best viewed original size.
Does anything there relate to MTF somehow "failing to work" for the Bayer?
You have my sympathy; been there and done that ...
OK.
My eyes are pretty poor, which is prefer numbers, graphs, just noticeable differences and so forth.
I agree!
Of course, correlation doesn't prove causation... it could be that the supposed subtle colour rendition of the SD15 was an artefact that just happens to look good to you. How could you tell the difference?
Pixel shift might give you the best of both worlds as it removes demosaicking.
Use a test original that shows coloured moire effects. Those colours are not caused by the lens but by the sensor.
Don
But do you get the fine detail on fast-moving subjects such as flowers that you can get with a Foveon sensor ?
Don
My G9 has something called Mode 2 which provides motion compensation. It worked when a jogger ran across my frame. I'm not sure if the price is a loss of detail in the moving area but it seems likely.
Perhaps I should have said 96%.An interesting measure.
However, most of the curves I see drop below 100% immediately above zero frequency. for example:
Image courtesy of Bob Atkins :
In some curves, usually for an over-sharpened image, the MTF rises above 100% then falls back below 100% at some frequency - so there could be two values for "detail" in such an image.
Still the basis of the idea is sound enough because, as implied, high contrast transfer at low detail is preferable for some folks. Hence my comment about the shape of the curve.
Perhaps.Perhaps I should have said 96%.An interesting measure.
However, most of the curves I see drop below 100% immediately above zero frequency. for example:
Image courtesy of Bob Atkins :
In some curves, usually for an over-sharpened image, the MTF rises above 100% then falls back below 100% at some frequency - so there could be two values for "detail" in such an image.
Still the basis of the idea is sound enough because, as implied, high contrast transfer at low detail is preferable for some folks. Hence my comment about the shape of the curve.
A bit like some histograms where a log X-axis is available ...
Which is probably why Grainger et al came up with the Subjective Quality Factor in the 70's which combines human vision with MTF, well-explained by Bob Atkins here:
Yes, that graph makes my point.Perhaps.Perhaps I should have said 96%.An interesting measure.
However, most of the curves I see drop below 100% immediately above zero frequency. for example:
Image courtesy of Bob Atkins :
In some curves, usually for an over-sharpened image, the MTF rises above 100% then falls back below 100% at some frequency - so there could be two values for "detail" in such an image.
Still the basis of the idea is sound enough because, as implied, high contrast transfer at low detail is preferable for some folks. Hence my comment about the shape of the curve.
A bit like some histograms where a log X-axis is available ...
Which is probably why Grainger et al came up with the Subjective Quality Factor in the 70's which combines human vision with MTF, well-explained by Bob Atkins here:
http://www.bobatkins.com/photography/technical/mtf/mtf4.html
Fortunately, QuickMTF also gives the linear frequency value to several significant figures. So, although it looks a bit cramped on the graph, your 96% comes out with the numbers 0.1 cy/px or 388.8 LPH.Yes, that graph makes my point.Perhaps.Perhaps I should have said 96%.An interesting measure.
However, most of the curves I see drop below 100% immediately above zero frequency. for example:
Image courtesy of Bob Atkins :
In some curves, usually for an over-sharpened image, the MTF rises above 100% then falls back below 100% at some frequency - so there could be two values for "detail" in such an image.
Still the basis of the idea is sound enough because, as implied, high contrast transfer at low detail is preferable for some folks. Hence my comment about the shape of the curve.
A bit like some histograms where a log X-axis is available ...
Which is probably why Grainger et al came up with the Subjective Quality Factor in the 70's which combines human vision with MTF, well-explained by Bob Atkins here:
http://www.bobatkins.com/photography/technical/mtf/mtf4.html
I think you need two graphs to see what's going on -- one with a linear X scale for fine detail, and one with a log scale to put it in proportion (and to show the effect of lens flare).
Yes I forgot this. In real life shooting though, I seldom have seen color moire. I think it is very rare in real life shooting.
Somewhat belated but this page from Jack Hogan aggrees with your feeling that lower frequencies are more important, especially when viewing an image at a proper distance:Yes, that graph makes my point.Perhaps.Perhaps I should have said 96%.An interesting measure.
However, most of the curves I see drop below 100% immediately above zero frequency. for example:
Image courtesy of Bob Atkins :
In some curves, usually for an over-sharpened image, the MTF rises above 100% then falls back below 100% at some frequency - so there could be two values for "detail" in such an image.
Still the basis of the idea is sound enough because, as implied, high contrast transfer at low detail is preferable for some folks. Hence my comment about the shape of the curve.
A bit like some histograms where a log X-axis is available ...
Which is probably why Grainger et al came up with the Subjective Quality Factor in the 70's which combines human vision with MTF, well-explained by Bob Atkins here:
http://www.bobatkins.com/photography/technical/mtf/mtf4.html
I think you need two graphs to see what's going on -- one with a linear X scale for fine detail, and one with a log scale to put it in proportion (and to show the effect of lens flare).
Don
I took a DP2 (not M) street shot and ran it through SPP4 with no adjustments. As expected there was some sharpening [by SPP] of the original slightly soft raw image.Somewhat belated but this page from Jack Hogan agrees with your feeling that lower frequencies are more important, especially when viewing an image at a proper distance:Yes, that graph makes my point.Perhaps.Perhaps I should have said 96%.An interesting measure.
However, most of the curves I see drop below 100% immediately above zero frequency. for example:
Image courtesy of Bob Atkins :
In some curves, usually for an over-sharpened image, the MTF rises above 100% then falls back below 100% at some frequency - so there could be two values for "detail" in such an image.
Still the basis of the idea is sound enough because, as implied, high contrast transfer at low detail is preferable for some folks. Hence my comment about the shape of the curve.
A bit like some histograms where a log X-axis is available ...
Which is probably why Grainger et al came up with the Subjective Quality Factor in the 70's which combines human vision with MTF, well-explained by Bob Atkins here:
http://www.bobatkins.com/photography/technical/mtf/mtf4.html
I think you need two graphs to see what's going on -- one with a linear X scale for fine detail, and one with a log scale to put it in proportion (and to show the effect of lens flare).
Don
https://www.strollswithmydog.com/mtf50-perceived-sharpness/
See under heading "MTF50 not relevant at Standard Distance" ...
As a Sigma camera user, I like the detail provided in the Sigma camera images. But the problem is that the proper level of "detail" complements (in a good way) the image concerned. It is meaningless on its own. Or worse still, an utter distraction unless primarily supporting "the image" itself. It's kind of fun to take a picture of a forest scene and the examine the leaves, etc. Fun photo images are...fun. Are they great? Few images are great. Detail is at most only part of what goes into making a good image.
